Agentic AI for Civic Navigation

I developed the full-stack AI architecture for a civic navigation assistant designed to guide users through complex municipal services. I engineered the entire experience, from real-time streaming responses to the underlying conversation logic that maintains context across multi-turn interactions.

The Challenge

Standard assistants cannot effectively navigate the non-linear, client-side rendered (SPA) structures of modern government platforms. I sought to bridge the gap between static retrieval and active, state-aware guidance within dynamic web environments.

The Solution

I built a specialized interface layer that allows the agent to perceive and interact with the DOM directly. By utilizing zero-shot calibration to interpret site hierarchies on-the-fly, the agent autonomously steers users through complex workflows without requiring manual site-mapping or pre-defined schemas.

Tech & Skills

• Core Competencies: Agentic Systems, Conversational AI, Web Automation, Zero-Shot AI